Temporal sequences, even after stationarization, often exhibit leptokurtic distributions with fat tails and persistent distribution shifts. These properties destabilize feature dynamics, amplify model variance, and hinder model convergence in time series forecasting. To address this, we propose Morphing-Flow (MoF), a framework that combines a spline-based transform layer (Flow) and a test-time-trained method (Morph), which adaptively normalizes non-stationary, fat-tailed distributions while preserving critical extreme features. MoF ensures that inputs remain within a network’s effective activation space—a structured, normal-like distribution—even under distributional drift. Experiments across eight datasets show that MoF achieves state-of-the-art performance: With a simple linear backbone architecture, it matches the performance of state-of-the-art models on datasets such as Electricity and ETTh2. When paired with a patch-based Mamba architecture, MoF outperforms its closest competitor by 6.3% on average and reduces forecasting errors in fat-tailed datasets such as Exchange by 21.7%. Moreover, MoF acts as a plug-and-play module, boosting performance in existing models without architectural changes.

Soft robots, compared to regular rigid robots, as their multiple segments with soft materials bring flexibility and compliance, have the advantages of safe interaction and dexterous operation in the environment. However, due to its characteristics of high dimensional, nonlinearity, time-varying nature, and infinite degree of freedom, it has been challenges in achieving precise and dynamic control such as trajectory tracking and position reaching. To address these challenges, we propose a framework of Deep Koopman-based Model Predictive Control (DK-MPC) for handling multi-segment soft robots. We first employ a deep learning approach with sampling data to approximate the Koopman operator, which therefore linearizes the high-dimensional nonlinear dynamics of the soft robots into a finite-dimensional linear representation. Secondly, this linearized model is utilized within a model predictive control framework to compute optimal control inputs that minimize the tracking error between the desired and actual state trajectories. The real-world experiments on the soft robot “Chordata” demonstrate that DK-MPC could achieve highprecision control, showing the potential of DK-MPC for future applications to soft robots. More visualization results can be found at https://pinkmoon-io.github.io/DKMPC/.

We present $\textbf{FlowRL}$, a novel framework for online reinforcement learning that integrates flow-based policy representation with Wasserstein-2-regularized optimization. We argue that in addition to training signals, enhancing the expressiveness of the policy class is crucial for the performance gains in RL. Flow-based generative models offer such potential, excelling at capturing complex, multimodal action distributions. However, their direct application in online RL is challenging due to a fundamental objective mismatch: standard flow training optimizes for static data imitation, while RL requires value-based policy optimization through a dynamic buffer, leading to difficult optimization landscapes. FlowRL first models policies via a state-dependent velocity field, generating actions through deterministic ODE integration from noise. We derive a constrained policy search objective that jointly maximizes Q through the flow polciy while bounding the Wasserstein-2 distance to a behavior-optimal policy implicitly derived from the replay buffer. This formulation effectively aligns the flow optimization with the RL objective, enabling efficient and value-aware policy learning despite the complexity of the policy class. Empirical evaluations on DMControl and Humanoidbench demonstrate that FlowRL achieves competitive performance in online reinforcement learning benchmarks.

The recent development of zero-shot reinforcement learning (RL) has opened a new avenue for learning pre-trained generalist policies that can adapt to arbitrary new tasks in a zero-shot manner. While the popular Forward-Backward representations (FB) and related methods have shown promise in zero-shot RL, we empirically found that their modeling lacks expressivity and that extrapolation errors caused by out-of-distribution (OOD) actions during offline learning sometimes lead to biased representations, ultimately resulting in suboptimal performance. To address these issues, we propose Behavior-REgularizEd Zero-shot RL with Expressivity enhancement (BREEZE), an upgraded FB-based framework that simultaneously enhances learning stability, policy extraction capability, and representation learning quality. BREEZE introduces behavioral regularization in zero-shot RL policy learning, transforming policy optimization into a stable in-sample learning paradigm. Additionally, BREEZE extracts the policy using a task-conditioned diffusion model, enabling the generation of high-quality and multimodal action distributions in zero-shot RL settings. Moreover, BREEZE employs expressive attention-based architectures for representation modeling to capture the complex relationships between environmental dynamics. Extensive experiments on ExORL and D4RL Kitchen demonstrate that BREEZE achieves the best or near-the-best performance while exhibiting superior robustness compared to prior offline zero-shot RL methods. The official implementation is available at: https://github.com/Whiterrrrr/BREEZE.

Hierarchical reinforcement learning (HRL) exhibits remarkable potential in addressing large-scale and long-horizon complex tasks. However, a fundamental challenge, which arises from the inherently entangled nature of hierarchical policies, has not been understood well, consequently compromising the training stability and exploration efficiency of HRL. In this article, we propose a novel HRL algorithm, high-level model approximation (HLMA), presenting both theoretical foundations and practical implementations. In HLMA, a Planner constructs an innovative high-level dynamic model to predict the -step transition of the Controller in a subtask. This allows for the estimation of the evolving performance of the Controller. At low level, we leverage the initial state of each subtask, transforming absolute states into relative deviations by a designed operator as Controller input. This approach facilitates the reuse of subtask domain knowledge, enhancing data efficiency. With this designed structure, we establish the local convergence of each component within HLMA and subsequently derive regret bounds to ensure global convergence. Abundant experiments conducted on complex locomotion and navigation tasks demonstrate that HLMA surpasses other state-of-the-art single-level RL and HRL algorithms in terms of sample efficiency and asymptotic performance. In addition, thorough ablation studies validate the effectiveness of each component of HLMA.

Visual reinforcement learning (RL) has shown promise in continuous control tasks. Despite its progress, current algorithms are still unsatisfactory in virtually every aspect of the performance such as sample efficiency, asymptotic performance, and their robustness to the choice of random seeds. In this paper, we identify a major shortcoming in existing visual RL methods that is the agents often exhibit sustained inactivity during early training, thereby limiting their ability to explore effectively. Expanding upon this crucial observation, we additionally unveil a significant correlation between the agents' inclination towards motorically inactive exploration and the absence of neuronal activity within their policy networks. To quantify this inactivity, we adopt dormant ratio as a metric to measure inactivity in the RL agent's network. Empirically, we also recognize that the dormant ratio can act as a standalone indicator of an agent's activity level, regardless of the received reward signals. Leveraging the aforementioned insights, we introduce DrM, a method that uses three core mechanisms to guide agents' exploration-exploitation trade-offs by actively minimizing the dormant ratio. Experiments demonstrate that DrM achieves significant improvements in sample efficiency and asymptotic performance with no broken seeds (76 seeds in total) across three continuous control benchmark environments, including DeepMind Control Suite, MetaWorld, and Adroit. Most importantly, DrM is the first model-free algorithm that consistently solves tasks in both the Dog and Manipulator domains from the DeepMind Control Suite as well as three dexterous hand manipulation tasks without demonstrations in Adroit, all based on pixel observations.

The varying significance of distinct primitive behaviors during the policy learning process has been overlooked by prior model-free RL algorithms. Leveraging this insight, we explore the causal relationship between different action dimensions and rewards to evaluate the significance of various primitive behaviors during training. We introduce a causality-aware entropy term that effectively identifies and prioritizes actions with high potential impacts for efficient exploration. Furthermore, to prevent excessive focus on specific primitive behaviors, we analyze the gradient dormancy phenomenon and introduce a dormancy-guided reset mechanism to further enhance the efficacy of our method. Our proposed algorithm, ACE: Off-policy Actor-critic with Causality-aware Entropy regularization, demonstrates a substantial performance advantage across 29 diverse continuous control tasks spanning 7 domains compared to model-free RL baselines, which underscores the effectiveness, versatility, and efficient sample efficiency of our approach. Benchmark results and videos are available at https://ace-rl.github.io/.

Learning high-quality $Q$-value functions plays a key role in the success of many modern off-policy deep reinforcement learning (RL) algorithms. Previous works primarily focus on addressing the value overestimation issue, an outcome of adopting function approximators and off-policy learning. Deviating from the common viewpoint, we observe that $Q$-values are often underestimated in the latter stage of the RL training process, potentially hindering policy learning and reducing sample efficiency. We find that such a long-neglected phenomenon is often related to the use of inferior actions from the current policy in Bellman updates as compared to the more optimal action samples in the replay buffer. We propose the Blended Exploitation and Exploration (BEE) operator, a simple yet effective approach that updates $Q$-value using both historical best-performing actions and the current policy. Based on BEE, the resulting practical algorithm BAC outperforms state-of-the-art methods in over 50 continuous control tasks and achieves strong performance in failure-prone scenarios and real-world robot tasks. Benchmark results and videos are available at https://jity16.github.io/BEE/.

Training reinforcement learning policies using environment interaction data collected from varying policies or dynamics presents a fundamental challenge. Existing works often overlook the distribution discrepancies induced by policy or dynamics shifts, or rely on specialized algorithms with task priors, thus often resulting in suboptimal policy performances and high learning variances. In this paper, we identify a unified strategy for online RL policy learning under diverse settings of policy and dynamics shifts: transition occupancy matching. In light of this, we introduce a surrogate policy learning objective by considering the transition occupancy discrepancies and then cast it into a tractable min-max optimization problem through dual reformulation. Our method, dubbed Occupancy-Matching Policy Optimization (OMPO), features a specialized actor-critic structure equipped with a distribution discriminator and a small-size local buffer. We conduct extensive experiments based on the OpenAI Gym, Meta-World, and Panda Robots environments, encompassing policy shifts under stationary and non-stationary dynamics, as well as domain adaption. The results demonstrate that OMPO outperforms the specialized baselines from different categories in all settings. We also find that OMPO exhibits particularly strong performance when combined with domain randomization, highlighting its potential in RL-based robotics applications.

Off-policy reinforcement learning (RL) has achieved notable success in tackling many complex real-world tasks, by leveraging previously collected data for policy learning. However, most existing off-policy RL algorithms fail to maximally exploit the information in the replay buffer, limiting sample efficiency and policy performance. In this work, we discover that concurrently training an offline RL policy based on the shared online replay buffer can sometimes outperform the original online learning policy, though the occurrence of such performance gains remains uncertain. This motivates a new possibility of harnessing the emergent outperforming offline optimal policy to improve online policy learning. Based on this insight, we present Offline-Boosted Actor-Critic (OBAC), a model-free online RL framework that elegantly identifies the outperforming offline policy through value comparison, and uses it as an adaptive constraint to guarantee stronger policy learning performance. Our experiments demonstrate that OBAC outperforms other popular model-free RL baselines and rivals advanced model-based RL methods in terms of sample efficiency and asymptotic performance across **53** tasks spanning **6** task suites.

Model Predictive Control (MPC) has exhibited remarkable capabilities in optimizing objectives and meeting constraints. However, the substantial computational burden associated with solving the Optimal Control Problem (OCP) at each triggering instant introduces significant delays between state sampling and control application. These delays limit the practicality of MPC in resource-constrained systems when engaging in complex tasks. The intuition to address this issue in this paper is that by predicting the successor state, the controller can solve the OCP one time step ahead of time thus avoiding the delay of the next action. To this end, we compute deviations between real and nominal system states, predicting forthcoming real states as initial conditions for the imminent OCP solution. Anticipatory computation stores optimal control based on current nominal states, thus mitigating the delay effects. Additionally, we establish an upper bound for linearization error, effectively linearizing the nonlinear system, reducing OCP complexity, and enhancing response speed. We provide empirical validation through two numerical simulations and corresponding real-world robot tasks, demonstrating significant performance improvements and augmented response speed (up to 90%) resulting from the seamless integration of our proposed approach compared to conventional time-triggered MPC strategies.

Designing and analyzing model-based RL (MBRL) algorithms with guaranteed monotonic improvement has been challenging, mainly due to the interdependence between policy optimization and model learning. Existing discrepancy bounds generally ignore the impacts of model shifts, and their corresponding algorithms are prone to degrade performance by drastic model updating. In this work, we first propose a novel and general theoretical scheme for a non-decreasing performance guarantee of MBRL. Our follow-up derived bounds reveal the relationship between model shifts and performance improvement. These discoveries encourage us to formulate a constrained lower-bound optimization problem to permit the monotonicity of MBRL. A further example demonstrates that learning models from a dynamically-varying number of explorations benefit the eventual returns. Motivated by these analyses, we design a simple but effective algorithm CMLO (Constrained Model-shift Lower-bound Optimization), by introducing an event-triggered mechanism that flexibly determines when to update the model. Experiments show that CMLO surpasses other state-of-the-art methods and produces a boost when various policy optimization methods are employed.